Active learning for automatic generation of narratives from numeric financial and supply chain data
Large enterprises compile and analyze large amounts of data on a daily basis. Typically, the collected raw data is processed by financial analysts to produce reports. Executive personnel use these reports to oversee the operations and make decisions based on the data. Some of the processing performed by can be easily automated by currently available computational tools. These tasks mostly make use of standard transformations on the raw data including visualizations and aggregate summaries. On the other hand, automating some of the manual processing requires more involved AI techniques. In our project, we aim to solve one of these harder to automate tasks. In fact, analyzing textual data using NLP is one of the standardized methods of data processing in modern software tools. However, vast majority of NLP methods primarily aim to analyze textual data, rather than generate meaningful narratives. Since generation of text is a domain dependent and non-trivial task, automated generation of narratives requires novel research to be useful to an enterprise environment. In this project, we focus on using numerical financial and supply chain data to generate useful textual reports that can be used in executive level companies. Upon successful completion of this project, financial analysts will spend less time on repetitive tasks and have more time to focus on reporting tasks requiring higher level data fusion skills.
Advancing sustainable aerodynamic solutions with improved modeling
Within current aerospace design, it is necessary to over-engineer features to ensure stability and safety under emergency conditions. It would be ideal to develop capabilities to reduce the size of large elements of commercial aircraft with reliable technologies that ensure safe operation under hazardous conditions. A key advantage of the synthetic jet is that no bulky air source and supply system is required to provide actuation to the flow. The planned changes to the aircraft structure that increase fuel economy and reduce weight will ensure the success of the economically important Canadian aerospace industry.
Big cardiac data
We propose to develop a cloud based image centered informatics system powered by newly developed big data analytics from our group for automatic diagnosis and prognosis of heart failure (HF). HF is a leading cause of morbidity and mortality in Ontario and around the world. There is no cure. Therefore early diagnosis and accurate prediction is very critical before the symptoms appears. However, previously lack of efficient image analytics tool has been the major pitfall to have an accurate prognosis and early diagnosis system. The system will greatly improve the clinical accuracy and enables accurate early diagnosis and prediction by intelligently analyzing all associated historical images and clinical reports. The final system will not only greatly reduce the sudden death and irreversible cardiac conditions, but also offers a great optimization of decision system in current healthcare. This project is based on strength built upon one successful SOSCIP project and two OCE projects.
Big data analytics for the maritime internet of things (IoT)
The Internet of Things (IoT) is an emerging phenomenon that enables ordinary devices to generate sensor data and interact with one another to improve daily life. The maritime world has not escaped to the influence of the IoT revolution. We are in the midst of a technological wave in which vessels are not the only ones carrying sensors (GPS or radar) anymore, but other maritime entities such as cranes, crates, boats, pickup trucks, etc. are being equipped with the same capabilities. This trend constitutes the backbone of the so-called Maritime Internet of Things (mIoT).
This project is about exploiting the tide of sensor data emitted by a myriad of maritime entities in order to improve both internal and collaborative processes of mIoT-related organizations; for instance, think of a Port Authority adjusting its berthing and unloading schedule upon receiving notice that a vessel has been delayed by harsh weather conditions. The challenge addressed by this research project is the generation of actionable intelligence for Decision Support using Big Data analytics. Actionable intelligence includes anomalies, alerts, threats, potential response generation, process refinement and other types of knowledge that improve the efficiency of a maritime-related organization and/or the manner in which it interacts with other similar organizations.
Computational high-throughput screening of catalyst materials for renewable fuel and feedstock generation
According to a World Energy Council Report, population growth and rising standards of living across the world will at least double global energy demand by 2050. Simultaneously, carbon dioxide emissions must be reduced significantly to prevent a catastrophic rise in global temperatures. Clean and abundant renewable energy sources are available; unfortunately, the intermittency of solar and wind power is a prevailing problem which is limiting the potential for widespread use. Our project seeks to address both of these issues through development of novel catalysts to electrochemically convert CO2 captured from power plants into fuels and other higher value chemical feedstocks using renewable electricity. This innovative strategy will (1) provide a long term storage solution by converting renewable electricity into a stable chemical fuel, (2) provide a means to intelligently recycle CO2 rather than storing it in deep underground aquifers, and (3) provide a cleaner and cheaper pathway for production of industrial chemical feedstocks and fuels. This could be a truly disruptive technology which would allow Canadian led manufacturing of high value chemicals and fuels in a low-cost and low-carbon fashion. Additionally, there are large benefits to Canada’s energy sector by facilitating the dispatchability of renewable power.
Continuous vital sign monitoring using intelligent bed sheet
Studio 1 labs developed wireless intelligent bed sheet patient monitoring system that continuously captures client vital signs. In collaboration with Dr. Laura Nicholson and York University’s Faculty of Health, Studio 1 labs will work with SOSCIP infrastructure to match vital signs with gold standards approved medical decides for the highest level of accuracy through clinical validation and scientific evidence. With millions of data points collected from each device to output clinical grade quality information continuously, AI solutions allow modeling to predict health emergencies and diseases. This contributes to efficient health monitoring solutions that are simple and effective for use by older adults and healthcare providers.
Design of OLED materials for manufacturing and improved product quality
Organic light emitting diodes (OLEDs) present a unique opportunity to produce thinner and more efficient lighting and displays. This will change the way we interact with light. The main barrier to mass adoption of OLEDs is the manufacturing process, due to the need for high throughput while maintaining nanoscale precision. High throughput operation requires materials that can undergo elevated temperature without decomposing. Our objective is to use computational chemistry to model innovative materials that can withstand these elevated temperatures while still providing high performing OLEDs. We will simulate targeted compounds using SOSCIP’s computer cluster examining properties relevant to OLED manufacturing processes. Promising materials will be synthesized and their properties experimentally measured then compared to the simulation results. The most promising materials will then be integrated into OLEDs and characterized by OTI Lumionics in their pilot scale manufacturing line located in Toronto, ON.
Distributed Deep Learning and Graph Analytics Using IBM Spectrum Computing Solutions
Deep learning is a popular machine learning technique and has been applied to many real-world problems, ranging from computer vision to natural language processing. In most cases deep learning outperformed previous work. However, training a deep neural network is very time-consuming, especially on big data. A popular solution is to distribute and parallel the training process across multiple machines. Indeed, the race is on to parallelize deep learning! Industry and academic research teams around the world are trying to make deep neural networks train as fast as possible on farms of GPU capable servers. We are working with our IBM partners to help develop advanced scheduling and messaging techniques for distributed deep learning. In addition, we will develop two real-world applications of distributed deep learning to demonstrate the efficiency and effectiveness of distributed deep learning. In one application, we address the video surveillance problem of tracking a moving target over a network of video cameras with partial or no overlaps in their coverage. We will use a deep learning approach to identify multiple pedestrians in each video frame, and a particle filter to track moving pedestrians. In the second application, we address the problem of fraud/intrusion detection. We will use graph-based detection that considers relationships between objects or individuals. Graph-based approaches are powerful because they do not operate on objects or individuals in isolation, but also consider their network information. We will emphasize on graph-based fraud detection methods that have a number of applications and potentially large impacts.
Efficient deep learning for real-time traffic event detection
Miovision is interested in designing the first affordable, low-power, energy efficient real time traffic event detection system that can be installed without the need to be powered by the grid nor the need to be connected directly to city installed infrastructure. Deep learning for traffic event detection can provide overwhelmingly superior accuracy and addresses most of the real-world scenarios that make competing detectors unsuitable for customer adoption. The challenge with deep learning is its complexity, which is currently infeasible for a self-powered real-world embedded detection system. Working with Dr. Alexander Wong and the Vision and Image Processing Lab at the University of Waterloo, the goal of this project is to develop technologies that can significantly reduce the complexity of deep learning for traffic event detection, while maintaining its accuracy and market fit, so that it can be deployed on a low-cost and low-powered hardware platform.
Electrochemical Fischer-Tropsch synthesis of renewable liquid fuels from CO2
Renewable electricity costs have been rapidly declining, enabling clean consumption of energy in many sectors. However, there is still demand for energy-dense liquid fuels, such as in heavy freight and air transportation. In this project, we will harness machine learning to develop technologies that enable the synthesis of liquid fuels from carbon dioxide and/or synthesis gas using renewable electricity. Industrially, liquid fuels can be synthesized from a mixture of carbon monoxide and hydrogen called synthesis gas (syngas). However, this process requires high temperatures and pressures, and is itself responsible for significant greenhouse gas emissions. We propose the use of electrocatalysis to produce these liquid fuels. To accomplish this, we will use computational modeling and machine learning methods to design electrocatalysts that efficiently convert CO2 or syngas into dense chemical fuels. These computational efforts will be validated through a parallel experimental approach that includes the fabrication of new catalyst formulations and the construction of prototype electrochemical flow cells. This project will enable the synthesis of clean, energy-dense liquid fuels that can replace the use of fossil-derived fuels in industry and transportation sectors. This project is a logical extension of the existing CO2 related projects previously underway with SOSCIP. This project will allow the research to achieve the next milestones in the overall goal to achieve beneficial conversion of CO2.
High fidelity simulations and low-order aero-acoustic modeling of engine test cells
The testing and certification of gas turbines demand the well-controlled environment provided by an engine test cell. The resonant acoustic coupling between the flow generated noise from the gas turbine exhaust and the engine test cell impacts the quality and reliability of the engine testing and certification. Predictive modeling of flow generated noise using high-fidelity numerical simulations is central to an a priori acoustic assessment and for the development of noise-mitigating designs. As part of this effort, the Multi-Physics Interaction Lab and University of Waterloo will numerically study, with the help of high fidelity, large-eddy simulations of the SOSCIP high-performance computers, the acoustic noise generation in partially confined jets undergoing a re-acceleration through the test cell ejector system. As a direct outcome, the researchers will develop a low order Aero-acoustic model that will be used by our industrial partner to predict resonant acoustic models to within +/- 20% the frequency and amplitude of the coupling phenomena. This OCE and NSERC-funding project will permit the training and mentoring of four HQP for careers in science and technology within Canada.
Large scale simulation of nanostructured optical surfaces
Plasmonic metasurfaces are engineered human-fabricated surfaces on which there is nanometer scale structure, usually containing metallic features. These nanoscale features can resonantly interact with light through the excitation of electric charge density oscillations, or surface plasmons. As plasmonic metasurfaces have essentially infinite design potential — including designer resonant behaviour and strong nanoscale light field enhancements — there is currently a great deal of interest in using them for a wide range of applications, from flat optical devices to biosensing to enhanced nonlinear optical signals3 to colour production4. A major focus of this project will be to perform large scale computational electrodynamics simulations on the Blue Gene Q to understand and design plasmonic metasurfaces. The work will be connected to projects underway with several Canadian industrial partners, as well as a plan to engage with several others, and will involve multiple trainees at various stages in their education – from undergraduate to postdoctoral fellow.
Mobilizing ‘Big energy data’ for building conservation to socialize sustainability
This project will establish the first customer-centric universal tool using mobile/cloud technology to deliver & share energy information electronically with energy customers to stimulate social change, encourage conservation & help manage/reduce GHG emissions.
A global desire to manage climate change is here. There needs to be a joint effort to help the energy user gain access to electronic information, so the data can be used to better manage conservation, sustainability and technology integration. Innovation in the energy sector continues to create data silos, as new technologies used don’t allow for the data to freely flow out of the systems to other solutions. Customers require easy access & a simplified understanding of electric information (not print or PDFs). For this to happen, the end customer requires data to be managed, simplified & shared, so information can be “open” to innovation and electronic data to be standardized.
The market needs Green practices and conservation to flourish in the private sector and public sector. Internationally and in Ontario, there are regulatory and political requirements to provide end-user access to electronic data, but without a simplified roadmap to interconnectivity how will it happen? The market needs linkages to the “Internet of Things” to learn, inform and conserve. In a few short years, there will be more than 25 billion devices generating data on every topic imaginable. The energy customer and building owner needs simplified data from multiple sources. Screaming has an opportunity to deliver this information cost-effectively to the utility, government & customer.
Multi-sensor miner data using smart computing platforms
The mining industry is a significant contributor to the Canadian economy, requiring close to 100,000 people to be hired by 2020. However, mining remains the second most dangerous job across the planet with 46.9 fatalities per 100,000 workers. Until the recent MINER act, there is little interest in providing communication services in mines. The MINER act made tracking all coal miners mandatory. In this project, we aim to develop a smart computing platform that combines data from digital wireless radio, thermal and video images in real time. This platform will incorporate tracking algorithms to aid in precisely locating miners, explosives and vehicles in mines continuously. This is very difficult due to the very nature of the mines. Extremely harsh, irregularly confined, and rough mine environments make radio wave propagation unpredictable. Furthermore, low visible light conditions prevent effective surveillance using cameras. Increasing number of sensors helps accurately locate targets, but will also increase the complexity of the algorithms. Using a variety of sensors will require several algorithms to run simultaneously to extract and fuse useful information from noisy data gathered from the mine in real-time. Especially identification of moving object in low-light video and thermal imaging is computationally very intensive. Hence, such a platform can only be implemented on a supercomputer, or a network of very fast computers. For this purpose, our work will rely on SOSCIP’s extensive computational capabilities. The result of this work will lead to significant improvements in safety of the mine personnel and better resource allocation and management in the mine
Next generation radio interferometry for astronomy, geodesy and radar ranging
Determining the properties of a pulsar and the interstellar medium through which its pulses of radio emission travel may appear to be an endeavor that is rather different from using a radar to localize a satellite and determine its properties. But both share the short bursts of radio emission and while in one case the pulses are refracted and in the other reflected, the analysis of the received radio emission turns out to require very similar techniques. Recognizing this synergy, this project launches a new dimension in the industrial academic collaboration between Thoth Technology Inc. and the University of Toronto, to jointly develop technology for using transient radio signals to localize sources and determine their properties, both to better understand pulsars and the interstellar medium and to enable Thoth to provide and commercialize unique radar capabilities. The project plans to first develop a common interface for the Thoth Algonquin Radio Observatory facilities, enabling the production of data compliant with the VDIF digital media standard, even at very high data rates, and to select and combine parts from the streams at will for rapid transmission and quick-look visualization. Next, we use BGQ resources to develop and optimize the data analysis, following new insights on how bursty signals in particular should allow simpler retrieval of properties of sources.
Using advanced analytics to develop a multimodal signature of concussion and postconcussive syndrome
Concussion are extremely common in deployment and in military and civilian activities (i.e.) sports). Persisting symptoms that make up “post-concussive syndrome” (PCS) including headaches, balance difficulties , depression and anxiety can occur in 10-15% of cases. The diagnosis of concussion and PCS is currently based on a patient’s report of their symptoms and a physical exam. Research, including our own, has explored the value of specific tests including those that use eye movements, neuropsychological tests and MRI. Although useful in the research setting, we do not understand the value of these tests when used together and need to know what aspects of those tests are most valuable in developing future tools that distinguish those who are injured from those who are not. For this research we will utilize a dataset collected over the last four years, that contains MRI, neuropsychological, eye movement and other data from concussed, PCS and non-injured individuals. We will apply machine learning based methods (convolutional and long short term memory neural networks) to analyze the data to define more sensitive and specific tests. These tools may be used in both a military and civilian setting allowing for more personalized treatment and recovery programs thereby lessening the burden of concussion and PCS.
Using brain fMRI machine learning as a predictor of PTSD treatment outcomes for Canadian Military
Approximately 13% of Canadian Armed Forces (CAF) members and veterans deployed to Afghanistan are diagnosed with a deployment-related mental disorder, such as post-traumatic stress disorder (PTSD) and many will experience comorbid major depressive disorder. The proposed study will utilize brain imaging data (fMRI) to determine if neurobiological machine learning algorithms can predict treatment outcomes and psychiatric symptomatology in CAF members and veterans. This research will benefit CAF members and veterans through the identification and clinical application of novel avenues to personalized medicine; namely, using neuroimaging data to predict probable treatment outcomes and aid in the selection of appropriate treatment methodologies. The proposed research requires making data linkages and generalizing datasets by combining imaging and clinical outcomes data; however, by overcoming these challenges, we anticipate developing a tool that can aid in the diagnosis of PTSD and its various subtypes as well as inform treatment guidelines.
Using machine learning to investigate sympathetic activation of the autonomic
The goal of this research is to further our understanding and clinical management of Canadian Forces service members and veterans suffering from a complex medical triad of traumatic brain injury, chronic pain, and post-traumatic stress disorder. Over half of rehabilitation patients experience one or more of these complex medical conditions, often associated with intractable symptoms which do not respond to traditional treatment options, and impairing their ability to function effectively at work and in the community. Using a Computer Assisted Rehabilitation Environment (CAREN) this research will collect and consolidate a series of non-invasive whole body biological measurements from patients during immersive therapy sessions in the CAREN Virtual Reality facility. High-performance computing and machine learning will be used to develop and deploy real-time estimators of SAANS. These systems will allow clinicians to create individualized treatment plans for patients, thereby maximizing rehabilitation benefits and avoiding costly setbacks in patient treatment.